Adaptive robotic tutors for scaffolding self-regulated learning

This thesis explores how to utilise social robotic tutors to tackle the problem of providing children with enough personalised scaffolding to develop Self-Regulated Learning (SRL) skills. SRL is an important 21st century skill and correlates with measures of academic performance. The dynamics of social interactions when human tutors are scaffolding SRL are modelled, a computational model for how these strategies can be personalised to the learner is developed, and a framework for long-term SRL guidance from an autonomous social robotic tutor is created. To support the scaffolding of SRL skills the robot uses an Open Learner Model (OLM) visualisation to highlight the developing skills or gaps in learners' knowledge. An OLM shows the learner's competency or skill level on a screen to help the learner reflect on their performance. The robot also supports the development of meta-cognitive planning or forethought by summarising the OLM content and giving feedback on learners' SRL skills. Both short and longer-term studies are presented, which show the benefits of fully autonomous adaptive robotic tutors for scaffolding SRL skills. These benefits include the learners reflecting more on their developing competencies and skills, greater adoption SRL processes, and increased learning gain.